Chloe Maraina brings a sharp, analytical perspective to the high-stakes world of enterprise data infrastructure. With a background rooted in Business Intelligence and a passion for uncovering the visual stories hidden within massive datasets, she helps organizations navigate the complex transition from static legacy systems to AI-ready architectures. Today, we dive into how enterprises can break free from vendor lock-in and build an open infrastructure that truly fuels the next generation of digital agents.
How do legacy licensing structures and shifting pricing models from major vendors create a bottleneck for organizations attempting to move toward AI readiness?
It is a frustrating reality where the conversation about innovation often dies in the procurement office before it even reaches the engineering floor. We are seeing major players like Oracle and Broadcom impose aggressive pricing shifts that effectively stall modernization efforts, forcing teams to burn their budgets on maintaining the status quo rather than investing in new capabilities. When licensing costs spike unexpectedly, the capital intended for experimental agentic AI projects is diverted just to keep the lights on. This financial weight creates a paralyzing atmosphere where leadership becomes risk-averse, hesitant to greenlight any project that doesn’t offer an immediate, guaranteed return. Ultimately, these legacy blockers prevent the structural agility required to even begin the journey toward a modern, AI-integrated data stack.
We often hear that moving to the cloud is not a silver bullet, and many teams find themselves stuck after a lift-and-shift migration. What are the specific structural failures that occur when data remains fragmented across separate systems?
The lift-and-shift approach is frequently a trap because it moves existing messes from an on-premise basement to a cloud-based penthouse without actually cleaning them up. Organizations find themselves drowning in data sprawl where transactions, analytics, and AI workloads live in entirely separate silos, leading to bloated costs and a terrifying amount of governance risk. When your data is fragmented this way, you are not just paying for multiple storage solutions; you are paying the “technical debt” of trying to synchronize them constantly. It leaves engineers feeling like they are fighting fires rather than building value, as the architecture lacks the cohesive skeleton needed for real-time analytics. Without a unified platform, the dream of a seamless data flow remains out of reach, leaving teams with high cloud bills and very little to show for it in terms of AI-driven insights.
With SaaS applications and proprietary ecosystems increasingly closing off data access, how does the concept of Open Data Infrastructure offer a way out?
We are witnessing a quiet war for data ownership where SaaS providers monetize the very information they generate on a customer’s behalf, effectively building walled gardens that make data extraction a nightmare. Open Data Infrastructure is the necessary rebellion against this vendor lock-in, emphasizing automated, standards-based ingestion and an open data lake foundation. It is about ensuring that your data remains portable and interoperable by design, so you can evolve your stack without the trauma of a total re-platforming every few years. By committing to open standards, an organization can finally achieve a unified activation layer where semantics and AI consumption are streamlined across the board. This flexibility is the only way to future-proof a strategy in an era where proprietary ecosystems are specifically designed to keep you trapped and paying premiums.
Gartner predicts that 40% of agentic AI projects risk cancellation by 2027. In your view, what is the primary disconnect between current data strategies and the requirements of successful AI agents?
That 40% figure is a sobering wake-up call for any executive who thinks they can simply “bolt on” AI to a broken foundation. The disconnect lies in the fact that agentic AI requires continuous, scalable workloads that the traditional modern data stack was never actually built to handle. Agents do not just need access to a periodic report; they need to live within a data environment that powers real-time operations and sophisticated, autonomous decision-making. When projects are stuck in the staging phase because the underlying data is too slow, too messy, or too expensive to access, the entire initiative loses momentum and eventually gets the axe. To avoid being part of that cancellation statistic, enterprises must shift their focus from static reporting to building a strategy specifically designed to scale across these intelligent agents.
Given the need for a sovereign and flexible platform, how does an open-source solution like EDB Postgres AI bridge the gap between on-premise stability and cloud scalability?
EDB Postgres AI is a significant development because it offers a truly sovereign platform that looks and feels the same whether it is deployed in the cloud, on-premise, or in a fully air-gapped environment. This consistency is vital for large enterprises that must maintain strict security protocols while still pursuing the cutting edge of AI-driven automation. As the leading contributor to the Postgres ecosystem, EDB has extended the solution to ensure that data now powers operations and real-time systems rather than just sitting in back-end archives. It provides a massive sense of relief for infrastructure teams who are tired of managing different versions of the same database across fragmented environments. By unifying these workloads, it reduces the friction of migration and allows the organization to focus on building intelligent applications rather than just managing silos.
What is your forecast for the evolution of the data stack over the next three years?
I predict a massive shift away from proprietary “all-in-one” clouds as the realization sinks in that data portability is the ultimate competitive advantage in the AI era. We will see the rise of “sovereign data” where companies reclaim ownership of their information from SaaS providers, leveraging open-source foundations to build their own custom AI ecosystems. The 40% of AI projects currently at risk of cancellation will either fail or be reborn as part of a more disciplined infrastructure that prioritizes real-time ingestion over old-fashioned batch processing. By 2027, the organizations that thrive will be those that treated their data architecture as a living, interoperable organism rather than a series of static, disconnected buckets. It is going to be a period of intense architectural hygiene where the winners are defined by the cleanliness and accessibility of their data.
